GitHub user WillAyd added a comment to the discussion: A new home for 
pyarrow-stubs?

I can speak a bit to the pandas perspective since that was brought up here. 
Pandas does have stubs with an entirely separate team that manages them (the 
owner of that repo is a pandas core team member). They aren't always up to 
date, but generally the team that manages them does a decent job of keeping 
them consistent with pandas releases. Typing issues are reported in the 
separate stubs repository and usually not noticed by pandas core teammembers.

The inline annotations are used for internal automated type checking 
validation, with the idea being that these would catch bugs. However, there are 
also plenty of ignores scattered throughout the code base, because pandas has 
worked on stubs for many years, dealt with different versions of mypy / pyright 
/ etc..., and been contributed by people with different levels of interest when 
it comes to type maintenance. So its questionable if that has actually helped 
on that front and been burdensome.

> It's both reducing the size and making maintenance (IMHO) easier.

The importance of this can't be stressed enough. We have had issues in the past 
with not having a clear directive on how important types are, so some 
contributors come in leaving them on the backburner and then feel like they get 
stonewalled during review to fix types. I do think its unfortunate to leave 
features on the table that don't satisfy the type system, but a clearer 
directive from a project team could probably avoid that issue.



GitHub link: 
https://github.com/apache/arrow/discussions/45919#discussioncomment-14296726

----
This is an automatically sent email for [email protected].
To unsubscribe, please send an email to: [email protected]

Reply via email to